Quick prediction method and system of proton exchange membrane fuel cell physical field cloud map

Abstract

The present application relates to the technical field of fuel cell digital design and intelligent modeling, and particularly relates to a quick prediction method of proton exchange membrane fuel cell physical field cloud chart, comprising the following steps: performing simulation calculation on a PEMFC model according to input information to obtain a continuous physical field cloud chart; converting the continuous physical field cloud chart into a discrete index label chart with limited color categories based on a global color palette; performing pixel-level classification prediction on the discrete index label chart based on a deep learning classification model to output a predicted physical field cloud chart. The predicted physical field cloud chart can be directly output under a new working condition, the dependence on repeated CFD simulation is reduced, quick prediction is realized, simulation calling frequency is reduced, and the design cycle is shortened.

Description

Technical Field

[0001] This invention belongs to the field of digital design and intelligent modeling technology for fuel cells, and particularly relates to a rapid prediction method and system for physical field cloud maps of proton exchange membrane fuel cells. Background Technology

[0002] The proton exchange membrane fuel cell (PEMFC) involves strongly coupled processes such as electrochemical reactions, heat and mass transfer, and water management. Its key physical fields (e.g., current density distribution, oxygen transport distribution, and membrane / catalyst layer water content distribution) directly determine the fuel cell's performance, uniformity, and durability. Currently, obtaining these physical fields mainly relies on computational fluid dynamics (CFD) simulations. However, CFD simulations are computationally intensive, often taking hours for a single simulation, which cannot meet the needs of real-time control and large-scale optimization design. Furthermore, high-precision simulations for multi-condition scanning require enormous computational resources, making rapid iteration difficult and increasing computational and time costs. Alternatively, training with traditional conditional generative adversarial networks (cGANs) is unstable, prone to pattern collapse or detail distortion with small sample sizes; while some models, in pursuit of high accuracy, often have deep networks with large numbers of parameters and long inference times, failing to meet the stringent millisecond-level prediction requirements of industrial environments.

[0003] Therefore, how to directly output predicted physical field cloud maps under new operating conditions, reduce reliance on repeated CFD simulations, achieve rapid prediction, reduce the frequency of simulation calls, and shorten the design cycle has become an urgent technical problem to be solved. Summary of the Invention

[0004] In view of this, the present invention aims to provide a rapid prediction method for the physical field distribution cloud map inside a proton exchange membrane fuel cell. Under new operating conditions, the predicted physical field cloud map can be directly output, reducing the reliance on repeated CFD simulations, achieving rapid prediction, reducing the frequency of simulation calls, and shortening the design cycle.

[0005] To achieve the above objectives, the technical solution created by this invention is implemented as follows: This invention provides a rapid prediction method for physical field cloud maps of proton exchange membrane fuel cells (PEMFCs), comprising the following steps: performing simulation calculations on a PEMFC model based on input information to obtain a continuous physical field cloud map; converting the continuous physical field cloud map into a discrete index label map with finite color categories based on a global color palette; performing pixel-level classification prediction on the discrete index label map based on a deep learning classification model, and outputting a predicted physical field cloud map.

[0006] Furthermore, the input information is operating condition parameters. Based on the input information, the PEMFC model is simulated and calculated to obtain a continuous physical field cloud map. Specifically, this includes: constructing the flow field form of the PEMFC model as a parallel flow field; and obtaining the continuous physical field cloud map by simulating and calculating the PEMFC model by changing the operating condition parameters.

[0007] Furthermore, the continuous physical field cloud map is transformed into a discrete index label map with a limited number of color categories based on the global color palette. Specifically, this includes: labeling each pixel in the continuous physical field cloud map with a color category index corresponding to the color category index of the global color palette; and forming the discrete index label map based on the color category index labels.

[0008] Furthermore, before converting the continuous physical field cloud map into a discrete index label map with a finite number of color categories based on the global color palette, the method further includes: performing image pixel clustering analysis on the CFD training data of the global CFD simulation image using the K-Means clustering algorithm to construct the global color palette, wherein the global color palette includes the pixels of the global simulation CFD image and the color category index corresponding to the finite number of categories.

[0009] Furthermore, before performing pixel-level classification prediction on the discrete index label map based on the deep learning classification model and outputting the predicted physical field cloud map, the method further includes: using a deep convolutional neural network (CNN) as the feature extractor of the deep learning classification model to construct a high-fidelity deep learning classification model Class-GAN; or using a minimal fully connected network (Tiny-FC) as the feature extractor of the deep learning classification model to construct an ultra-high-speed deep learning classification model Lite-GAN.

[0010] Furthermore, a high-fidelity deep learning classification model, Class-GAN, is constructed, specifically including: the output layer of the high-fidelity deep learning classification model uses an activation function to predict the color category probability of each pixel in the discrete index label image; and / or, the output layer of the high-fidelity deep learning classification model uses a classification cross-entropy loss function to make a forced judgment on the color category probability output for each pixel.

[0011] Furthermore, the discrete index label map is classified and predicted at the pixel level based on a deep learning classification model to output a predicted physical field cloud map. Specifically, this includes: using the high-fidelity deep learning classification model or the ultra-high-speed deep learning classification model to perform pixel-level classification and prediction on the discrete index label map to obtain the predicted physical field cloud map.

[0012] In addition, this application also provides a rapid prediction system for physical field cloud maps of proton exchange membrane fuel cells, comprising: a continuous physical field cloud map processing module, used to perform simulation calculations on a PEMFC model based on input information to obtain a continuous physical field cloud map; a discrete index label map conversion module, used to convert the continuous physical field cloud map into a discrete index label map with a finite number of color categories based on a global color palette; and a pixel-level classification prediction module, used to perform pixel-level classification prediction on the discrete index label map based on a deep learning classification model and output a predicted physical field cloud map.

[0013] Furthermore, the present invention also provides an electronic device, comprising: a processor, a memory, and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the memory via the bus, and the machine-readable instructions are executed by the processor to perform the steps of the fast prediction method described above.

[0014] Furthermore, the present invention also provides a non-transitory computer-readable storage medium storing computer instructions, wherein a computer program is stored thereon, wherein the program is executed by a processor to perform the fast prediction method as described above.

[0015] Compared with existing technologies, the present invention provides a rapid prediction method for the physical field cloud map of a proton exchange membrane fuel cell (PEMFC), comprising the following steps: performing simulation calculations on a PEMFC model based on input information to obtain a continuous physical field cloud map; converting the continuous physical field cloud map into a discrete index label map with finite color categories based on a global color palette; and performing pixel-level classification prediction on the discrete index label map based on a deep learning classification model to output the predicted physical field cloud map. It can be seen that a precise mapping framework between fuel cell operating conditions and predicted physical field cloud maps is constructed through a deep learning classification model. By performing cluster analysis on the pixels of the global CFD simulation image, a global color palette with finite color categories is constructed, transforming the continuous physical field cloud map values ​​into discrete index label maps with finite color categories. This transforms the original continuous regression problem into a pixel-level classification problem, reducing the prediction difficulty and enhancing the boundary representation ability from the data representation level. By performing pixel-level classification prediction on the discrete index label map through a deep learning classification model, the predicted physical field cloud map can be directly output under new operating conditions, thereby reducing the dependence on repeated CFD simulations, achieving rapid prediction, reducing the frequency of simulation calls, and shortening the design cycle. It can perform finite offline simulation calculations of continuous physical field cloud maps, and quickly output multi-physics field distribution cloud maps under new operating conditions, reducing the frequency of simulation calls and shortening the design cycle. Attached Figure Description

[0016] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments and descriptions of the invention are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 A schematic flowchart of a method for rapidly predicting the internal physical field distribution cloud map of a proton exchange membrane fuel cell is provided for embodiments of the present invention. Figure 2 A schematic diagram comparing the prediction paths of the fast prediction method provided in this embodiment of the invention with those of traditional prediction methods; Figure 3 This is a schematic diagram illustrating the transformation of the discrete index label map in the fast prediction method provided in this embodiment of the invention; Figure 4 A schematic diagram illustrating the construction of a high-fidelity deep learning classification model (Class-GAN) in the fast prediction method provided in this embodiment of the invention; Figure 5 A schematic diagram illustrating the construction of the ultra-high-speed deep learning classification model (Lite-GAN) in the fast prediction method provided in this embodiment of the invention; Figure 6 This is a schematic diagram illustrating the construction of a classification model using a traditional prediction method (cGAN). Figure 7 A schematic diagram illustrating the model performance analysis of the fast prediction method provided in this embodiment of the invention and the traditional prediction method; Figure 8 A schematic diagram of the continuous physical field cloud map and its corresponding CFD-Post color mark in the fast prediction method provided in the embodiments of the present invention; Figure 9 This is a schematic diagram of the global color palette in the fast prediction method provided in this embodiment of the invention; Figure 10 This is a schematic diagram of the structure of a computer device provided for an embodiment of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only for explaining the invention and do not constitute a limitation thereof. Similar elements in different embodiments are referred to by associated similar element reference numerals. In the following embodiments, many details are described to facilitate a better understanding of the invention. However, those skilled in the art will readily recognize that some features may be omitted in different situations, or may be replaced by other elements, materials, or methods. In some cases, some operations related to the invention are not shown or described in the specification. This is to avoid obscuring the core parts of the invention with excessive description. For those skilled in the art, detailed description of these related operations is not necessary; they can fully understand the related operations based on the description in the specification and general technical knowledge in the art.

[0018] It should be noted that, unless otherwise specified, the embodiments and features described in this invention can be combined to form various implementations. Furthermore, the order of the steps or actions in the method description can be changed or adjusted in a manner readily apparent to those skilled in the art. Therefore, the various orders in the specification and drawings are merely for the clear description of a particular embodiment and do not imply a mandatory order, unless otherwise stated that a particular order must be followed.

[0019] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation on this invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, features defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0020] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0021] The rapid prediction method for the physical field cloud map of a proton exchange membrane fuel cell provided in this invention application can directly output the predicted physical field cloud map under new operating conditions, reducing reliance on repeated CFD simulations, achieving rapid prediction, reducing the frequency of simulation calls, and shortening the design cycle. It is understood that the rapid prediction method provided in this invention application can also be applied to other types of fuel cells. The rapid prediction method and its steps provided in this invention application will be described in detail below.

[0022] Example 1 The invention will now be described in detail with reference to the accompanying drawings and embodiments. Figure 1 As shown, this invention provides a rapid prediction method for the physical field cloud map of a proton exchange membrane fuel cell (PEMFC), comprising the following steps: Step S10: Simulate the PEMFC model based on the input information to obtain a continuous physical field cloud map. The input information here can be fuel cell operating parameters such as temperature, humidity, stoichiometry, etc., or other external input conditions. Before the simulation calculation, the input operating parameters can be normalized and mapped to the [0,1] interval. The continuous physical field cloud map here is a high-dimensional spatial distribution cloud map, see [link to relevant documentation]. Figure 8 .

[0023] like Figure 1 , Figure 3 As shown, step S20: Based on the global color palette, the continuous physical field cloud map is transformed into a discrete indexed label map with a finite number of color categories. See [link to relevant documentation]. Figure 8 . Figure 8 This diagram illustrates the continuous physics field contour map and its corresponding CFD-Post color labels. The left side of the diagram shows the continuous physics field contour map, and the right side shows the corresponding CFD-Post scale map. A global color palette is used to accurately map the continuous physics field contour map into a discrete index label map with a finite number of color categories, thus obtaining the predicted physics field contour maps of the fuel cell under different operating conditions, displayed using color labels. CFD-Post is a CFD simulation post-processing software. Figure 8 As can be seen, the CFD-Post scale reflects a continuous correspondence between "color" and "physical value," primarily used to interpret the physical meaning of contour plots. Figure 9The global color palette shows that the discrete index label map reflects the discrete representation relationship of "color category - index number". It is mainly used to transform the continuous physical field prediction problem into a pixel-level classification problem, which is convenient for the subsequent training of deep learning models.

[0024] Figure 9 This is a global color palette derived from pixel clustering of global CFD simulation images. Transforming continuous physics field cloud maps into discrete indexed label maps transforms the continuous numerical regression problem into a pixel-level classification problem with a finite number of classes, reducing learning difficulty and improving training stability. Furthermore, using a unified global color palette constrains the consistency of color representation across different samples, reducing color drift and prediction bias. Therefore, during data preprocessing, the pre-built global color palette transforms physics field prediction into a pixel-level color classification problem.

[0025] Step S30: Perform pixel-level classification prediction on the discrete index label map based on a deep learning classification model, and output the predicted physics cloud map. This invention proposes a physics cloud map prediction framework based on a discretization classification strategy. The discrete index label map representation is more conducive to supervised training using Softmax and cross-entropy loss, thereby improving boundary sharpness and structural detail recovery capabilities. In physics boundaries and gradient abrupt change regions, pixel-level classification prediction is more likely to form clear edges than continuous regression, thus reducing blurring and smoothing phenomena.

[0026] Furthermore, in the rapid prediction method provided in this embodiment of the invention, step S10: the input information is the operating condition parameters, and the PEMFC model is simulated and calculated according to the input information to obtain a continuous physical field cloud map, specifically including: constructing the flow field form of the PEMFC model as a parallel flow field; and simulating and calculating the PEMFC model by changing the operating condition parameters to obtain a continuous physical field cloud map.

[0027] This step involves data collection and preprocessing of the input information. For example, the PEMFC model can be constructed using the computational fluid dynamics software Fluent, resulting in a parallel flow field. By changing five key operating parameters—such as cathode / anode relative humidity, cathode / anode stoichiometry, and operating temperature—167 sets of simulation calculations can be performed to obtain the corresponding steady-state continuous physical field cloud map. This physical field cloud map includes current density maps, water content distribution maps, and oxygen concentration maps. It should be noted that the selection of these 167 operating parameters can be achieved through orthogonal experiments, Sobol sequences, adaptive sampling, etc.

[0028] Furthermore, in the fast prediction method provided in this embodiment of the invention, step S20: converting the continuous physical field cloud map into a discrete index label map with a finite number of color categories based on a global color palette, specifically includes: labeling each pixel in the continuous physical field cloud map with a color category index corresponding to the color category index of the global color palette; and forming a discrete index label map based on the color category index labels.

[0029] Further, before step S20: converting the continuous physical field cloud map into a discrete index label map with finite color categories based on the global color palette, the fast prediction method provided in this embodiment of the invention further includes: performing image pixel clustering analysis on the CFD training data of the global CFD simulation image using the K-Means clustering algorithm to construct a global color palette, which includes the pixels of the global simulation CFD image and the corresponding finite category color category indexes. For example... Figure 3 As shown, the method of processing continuous physics cloud maps using the K-Means clustering algorithm can perfectly extract the colors of continuous physics cloud maps, avoiding subsequent color deviation problems.

[0030] Global color quantization is performed based on the global color palette constructed above. The K-Means clustering algorithm is used to cluster the RGB pixels of the limited CFD training data, constructing a global color palette containing multiple (e.g., 32) representative color category indices. Then, as follows... Figure 3 As shown, all continuous physical field cloud maps are transformed into discrete indexed label maps (LabelMap). The value of each pixel is no longer a physical quantity, but a color category index (0~31) in the global color palette, with a discrete interval of 32. The number of probability categories for the color category index can be adjusted: the discrete interval N is not limited to 32, and can be adjusted to 64 or 128 according to the accuracy requirements, or an adaptive interval division can be used to provide a data foundation for high-fidelity prediction.

[0031] Further, before step S30: performing pixel-level classification prediction on the discrete index label map based on the deep learning classification model and outputting the predicted physical field cloud map, the fast prediction method provided in this embodiment of the invention further includes: step S40: using a deep convolutional neural network (CNN) as the feature extractor of the deep learning classification model to construct a high-fidelity deep learning classification model, Class-GAN. Figure 4 As shown, the fully connected layers in Class-GAN can be replaced by ResNet residual networks or Attention mechanisms to further extract deeper features of the operating parameters.

[0032] The key feature of the Class-GAN architecture is its use of a deep convolutional neural network (CNN) as a feature extractor. Leveraging the local perception capabilities of convolutional layers, it maximizes the reproduction of minute textures and structural details within the physical field of a fuel cell. Compared to the blurry prediction performance of cGAN, Class-GAN effectively avoids the problem of blurred image boundaries through its classification mechanism, significantly improving the sharpness of physical field boundaries. It can accurately capture subtle differences between the flow channel and the ridge, achieving clear predictions.

[0033] In some embodiments, further, before step S30: performing pixel-level classification prediction on the discrete index label map based on the deep learning classification model and outputting the predicted physical field cloud map, the fast prediction method provided in the embodiments of this invention further includes: step S50: using the minimally oriented fully connected network Tiny-FC as the feature extractor of the deep learning classification model to construct the ultra-high-speed deep learning classification model Lite-GAN. Figure 5 As shown, Lite-GAN's architecture employs a minimalist fully connected network, eliminating all convolutional operations. Lite-GAN completely removes convolutional layers, utilizing only fully connected layers to directly establish a nonlinear mapping from operational parameters to the full-size physical field pixel space, achieving real-time prediction in low-computing-power environments. Compared to cGAN, Lite-GAN's inference speed is several times faster, reaching millisecond / microsecond levels, and significantly reduces memory usage, making it possible to deploy deep learning prediction models on resource-constrained industrial control chips.

[0034] Although the architecture of Lite-GAN is simplified, it still maintains sharp boundary features (rather than the fuzzy features of regression models) in its prediction results because it adopts the "quantization classification" method of this invention. At the same time, its inference speed is 1-2 orders of magnitude faster than convolutional networks, making it suitable for embedded deployment.

[0035] Furthermore, in the fast prediction method provided in this embodiment of the invention, step S40: constructing a high-fidelity deep learning classification model specifically includes: step S400: the output layer of the high-fidelity deep learning classification model uses an activation function to predict the color category probability of each pixel in the discrete index label image. The generator output layer of Class-GAN uses a Softmax activation function to predict the color category probability of each pixel in the discrete index label image, and then uses the classification cross-entropy loss function for guided training, overcoming the problem of blurred physical field boundaries caused by traditional regression methods.

[0036] Furthermore, in the fast prediction method provided in this embodiment of the invention, step S40: constructing a high-fidelity deep learning classification model specifically includes: the output layer of the high-fidelity deep learning classification model uses the classification cross-entropy loss function to force a judgment on the color category probability output by each pixel. Class-GAN abandons the traditional numerical regression approach and uses the global color palette constructed in the data preprocessing stage to transform the problem of predicting continuous physical fields into a pixel classification problem. By utilizing the characteristics of the classification cross-entropy loss function, a forced judgment is made within a clear physical value range, thereby eliminating image blur and generating a high-fidelity predicted physical quantity cloud map with sharp edges and rich texture details.

[0037] Furthermore, the fast prediction method provided in this embodiment of the invention includes step S30: performing pixel-level classification prediction on the discrete index label map based on a deep learning classification model and outputting a predicted physical field cloud map. Specifically, this includes: using a high-fidelity deep learning classification model or an ultra-high-speed deep learning classification model to perform pixel-level classification prediction on the discrete index label map to obtain a predicted physical field cloud map.

[0038] like Figure 2 As shown, the fast prediction method provided in the embodiments of this invention uses two independent deep learning classification models, Class-GAN and Lite-GAN. Users can choose one of the deep learning classification models to perform the prediction task according to the specific application scenario (whether it is high-fidelity analysis or real-time control).

[0039] Both Class-GAN and Lite-GAN output layers are designed with N channels (N=32) and use the Softmax activation function. The cross-entropy loss function guides training. Deep learning classification models predict the probability of each pixel belonging to one of the 32 colors at the pixel level, rather than the specific physical value.

[0040] The two different deep learning classification models, Class-GAN and Lite-GAN, used in the fast prediction method provided in this invention embodiment are taken as specific examples, along with the classification model cGAN using the traditional regression method (see modeling). Figure 5 The models are compared (as shown in the image). To fairly evaluate the performance of different modeling strategies, the outputs of all models are ultimately mapped back to the original physical numerical space or RGB color space for unified quantification calculation. The modeling CPU configuration is Intel(R) Core(TM) i7-13700F, GPU is NVIDIA RTX4060, deep learning library is PyTorch 2.0.1+cu117, Python version is 3.9.19, and ANSYS version is 2022R1.

[0041] To verify the effectiveness of the fast prediction method provided in this embodiment of the invention, the cGAN model is set as follows: Data processing: Traditional numerical regression method is used (normalized to continuous values ​​in [-1,1]), without color quantization.

[0042] Output mechanism: The output layer uses the Tanh activation function to predict continuous values.

[0043] Loss function: Use mean squared error (MSE) or L1 loss.

[0044] Limitations: Experimental results show that the cloud maps generated by this method exhibit obvious blurring and smoothing effects in areas with drastic gradient changes, such as the flow channel edge, and cannot achieve the clarity of Class-GAN / Lite-GAN used in the fast prediction method provided in this invention application.

[0045] As shown in Tables 1, 2, and 3, and Figure 7 The diagram illustrates the performance differences between the two models proposed in the fast prediction method of this invention and cGAN, using both quantitative and qualitative analysis. The quantitative analysis employs five metrics for evaluation: Mean Squared Error (MSE), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Structural Similarity Index (SSIM), and Peak Signal-to-Noise Ratio (PSNR). It can be seen that the fast prediction method provided in this invention transforms continuous physics field prediction from numerical regression of continuous physics field cloud maps to classification of discrete indexed label maps. Class-GAN achieves sharp imaging of the predicted physics field cloud map boundaries, while Lite-GAN enables deconvolutional and extremely fast inference, thus addressing to some extent the pain point of existing deep learning models struggling to balance accuracy and speed in fuel cell simulations. In the training of Lite-GAN, in addition to MSE loss, physical constraint loss (such as mass conservation loss) can be introduced into the loss function to improve the accuracy of the deep learning classification model without increasing the computational load of inference.

[0046] Table 1. Quantitative analysis of water content using different models

[0047] Table 2. Quantitative analysis of current density using different models

[0048] Table 3. Quantitative analysis of oxygen mass fraction using different models

[0049] Example 2 This invention application provides a rapid prediction system for the internal physical field distribution cloud map of a proton exchange membrane fuel cell, comprising: The continuous physics field cloud map processing module is used to perform simulation calculations on the PEMFC model based on input information to obtain a continuous physics field cloud map. The input information can be fuel cell operating parameters such as temperature, humidity, and stoichiometry, or other external input conditions. Before the simulation calculation, the input operating parameters can be normalized and mapped to the [0,1] interval. The continuous physics field cloud map here is a high-dimensional spatial distribution cloud map, such as... Figure 8 As shown. Figure 8 This diagram illustrates the continuous physics field contour map and its corresponding CFD-Post color labels. The left side of the diagram shows the continuous physics field contour map, and the right side shows the corresponding CFD-Post scale map. A global color palette is used to accurately map the continuous physics field contour map into a discrete index label map with a finite number of color categories, thus obtaining the predicted physics field contour maps of the fuel cell under different operating conditions, displayed using color labels. CFD-Post is a CFD simulation post-processing software. Figure 8 As can be seen, the CFD-Post scale reflects a continuous correspondence between "color" and "physical value," primarily used to interpret the physical meaning of contour plots. Figure 9 The global color palette shows that the discrete index label map reflects the discrete representation relationship of "color category - index number". It is mainly used to transform the continuous physical field prediction problem into a pixel-level classification problem, which is convenient for the subsequent training of deep learning models.

[0050] The discrete index label map conversion module is used to convert continuous physical field cloud maps into discrete index label maps with a limited number of color categories based on a global color palette. Figure 9 This is a global color palette derived from pixel clustering of global CFD simulation images. Transforming continuous physics field cloud maps into discrete indexed label maps transforms the continuous numerical regression problem into a pixel-level classification problem with a finite number of classes, reducing learning difficulty and improving training stability. Furthermore, using a unified global color palette constrains the consistency of color representation across different samples, reducing color drift and prediction bias. Therefore, during data preprocessing, the pre-built global color palette transforms physics field prediction into a pixel-level color classification problem.

[0051] A pixel-level classification prediction module is used to perform pixel-level classification prediction on discrete index label maps based on a deep learning classification model, outputting a predicted physics cloud map. This invention proposes a physics cloud map prediction framework based on a discretization classification strategy. Discrete index label map representation is more conducive to supervised training using Softmax and cross-entropy loss, thereby improving boundary sharpness and structural detail recovery capabilities. In physics boundaries and gradient abrupt change regions, pixel-level classification prediction is more likely to form clear edges than continuous regression, thus reducing blurring and smoothing phenomena.

[0052] Example 3 In addition, embodiments of the present invention also provide a computer device, a readable storage medium, and a computer program product.

[0053] Figure 10 This is a schematic diagram of the structure of a computer device 12 provided in an embodiment of the present invention. Figure 10 A block diagram of an exemplary computer device 12 suitable for implementing embodiments of the present invention is shown. Figure 10 The computer device 12 shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of the present invention.

[0054] like Figure 10 As shown, computer device 12 is represented in the form of a general-purpose computing device. Computer device 12 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0055] The components of the computer device 12 may include, but are not limited to: one or more processors or processing units 16, system memory 28, and bus 18 connecting different system components (including system memory 28 and processing unit 16).

[0056] Bus 18 represents one or more of several bus architectures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus architectures. For example, these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0057] Computer device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by computer device 12, including volatile and non-volatile media, removable and non-removable media.

[0058] System memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache 32. Computer device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 may be used to read and write non-removable, non-volatile magnetic media (…). Figure 10 Not shown; usually referred to as a "hard drive"). Although Figure 10 Not shown, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk") and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 18 via one or more data media interfaces. System memory 28 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of the present invention.

[0059] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in system memory 28. Such program modules 42 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. Program modules 42 typically perform the functions and / or methods described in the embodiments of the present invention.

[0060] Computer device 12 can also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), and with one or more devices that enable a user to interact with computer device 12, and / or with any device that enables computer device 12 to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed via input / output (I / O) interface 22. Furthermore, computer device 12 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 20. As shown, network adapter 20 communicates with other modules of computer device 12 via bus 18. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with computer device 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0061] The processing unit 16 executes various functional applications and data processing by running programs stored in the system memory 28, such as implementing the fast prediction method provided in the embodiments of the present invention.

[0062] This invention also provides a non-transitory computer-readable storage medium storing computer instructions, on which a computer program is stored, wherein the program, when executed by a processor, implements the fast prediction method provided in all embodiments of this invention.

[0063] The computer storage medium of this invention can be any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. More specific examples (a non-exhaustive list) of computer-readable storage media include: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0064] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0065] The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof. The computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0066] This invention also provides a computer program product, including a computer program that, when executed by a processor, implements the fast prediction method described above.

[0067] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this invention disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this invention can be achieved, and this is not limited herein.

[0068] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A rapid prediction method for physical field cloud maps of proton exchange membrane fuel cells, characterized in that: Includes the following steps: The PEMFC model is simulated based on the input information to obtain a continuous physical field cloud map. The K-Means clustering algorithm is used to perform image pixel clustering analysis on the CFD training data of global CFD simulation images to construct a global color palette. The global color palette includes the pixels of the global CFD simulation images and the color category indexes of the corresponding finite categories. The continuous physical field cloud map is transformed into a discrete indexed label map with a finite number of color categories based on a global color palette, specifically including: Corresponding to the color category index of the global color palette, each pixel in the continuous physical field cloud map is labeled with a color category index label; The discrete index label map is formed based on the color category index labels; The discrete index label map is classified and predicted at the pixel level based on a deep learning classification model, and the predicted physical field cloud map is output.

2. The rapid prediction method according to claim 1, characterized in that: The input information consists of operating parameters. Based on this input information, the PEMFC model is simulated to obtain a continuous physical field cloud map, specifically including: The flow field in the PEMFC model is a parallel flow field. The continuous physical field cloud map is obtained by performing simulation calculations on the PEMFC model by changing the operating parameters.

3. The rapid prediction method according to claim 1, characterized in that: Before performing pixel-level classification prediction on the discrete index label map based on a deep learning classification model and outputting the predicted physical field cloud map, the method further includes: A high-fidelity deep learning classification model, Class-GAN, is constructed by using a deep convolutional neural network (CNN) as the feature extractor of the deep learning classification model. Alternatively, a minimal fully connected network, Tiny-FC, can be used as the feature extractor for the deep learning classification model to construct an ultra-high-speed deep learning classification model, Lite-GAN.

4. The rapid prediction method according to claim 3, characterized in that: Building a high-fidelity deep learning classification model, Class-GAN, specifically includes: The output layer of the high-fidelity deep learning classification model uses an activation function to predict the color category probability of each pixel in the discrete index label image; and / or The output layer of the high-fidelity deep learning classification model uses the classification cross-entropy loss function to force a judgment on the color category probability of each pixel.

5. The rapid prediction method according to claim 3, characterized in that: The discrete index label map is classified and predicted at the pixel level based on a deep learning classification model, and the predicted physical field cloud map is output, specifically including: A high-fidelity deep learning classification model or an ultra-high-speed deep learning classification model is used to perform pixel-level classification prediction on the discrete index label map to obtain the predicted physical field cloud map.

6. A rapid prediction system for the internal physical field distribution cloud map of a proton exchange membrane fuel cell, characterized in that: include: The continuous physics field cloud map processing module is used to perform simulation calculations on the PEMFC model based on the input information to obtain a continuous physics field cloud map. The K-Means clustering algorithm is used to perform image pixel clustering analysis on the CFD training data of global CFD simulation images to construct a global color palette. The global color palette includes the pixels of the global CFD simulation images and the color category indexes of the corresponding finite categories. The discrete index label map conversion module is used to convert the continuous physical field cloud map into a discrete index label map with a finite number of color categories based on a global color palette. Specifically, it includes: Corresponding to the color category index of the global color palette, each pixel in the continuous physical field cloud map is labeled with a color category index label; The discrete index label map is formed based on the color category index labels; The pixel-level classification prediction module is used to perform pixel-level classification prediction on the discrete index label map based on a deep learning classification model, and output the predicted physical field cloud map.

7. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. The machine-readable instructions are executed by the processor to perform the steps of the fast prediction method as described in any one of claims 1 to 5.

8. A non-transitory computer-readable storage medium storing computer instructions, wherein a computer program is stored thereon, The program is executed by the processor using the fast prediction method as described in any one of claims 1 to 5.

Images